Overview of Recommendation Techniquesin Business Process Modeling?

Krzysztof Kluza, Mateusz Baran, Szymon Bobek, Grzegorz J. Nalepa

AGH University of Science and Technology,al. A. Mickiewicza 30, 30-059 Krakow, Poland

{kluza,matb,s.bobek,gjn}@agh.edu.pl

Abstract Modeling business processes is an important issue in Business ProcessManagement. As model repositories often contain similar or related models, theyshould be used when modeling new processes. The goal of this paper is to pro-vide an overview of recommendation possibilities for business process models.We introduce a categorization and give examples of recommendation approaches.For these approaches, we present several machine learning methods which can beused for recommending features of business process models.

1 Introduction

Business Process (BP) models are visual representations of processes in an organiza-tion. Such models can help to manage process complexity and are also easy to un-derstand for non-business user. Although there are many new tools and methodologieswhich support process modeling, especially using Business Process Model and Nota-tion (BPMN) [1], they do not support recommendation mechanisms for BP modelers.

As BPMN specifies only a notation, there can be several ways of using it. There arestyle directions how to model BPs [2], or guidelines for analysts based on BPs under-standability (e.g. [3]). However, a proper business process modeling is still a challeng-ing task, especially for inexperienced users.

Recommendation methods in BP modeling can address this problem. Based on cur-rent progress or additional pieces of information, various features can be recommendedto a modeler, and he/she can be assisted during designing models. Such assistance canprovide autocompleting mechanisms with capabilities of choosing next process frag-ments from suggested ones. Names of model elements or attachments can be recom-mended as well. Such approaches can reduce number of errors during process design aswell as speed up modeling process. It also supports reusing of existing process models,especially when a process repository is provided.

The rest of this paper is organized as follows: In Section 2, we provide a cate-gorization of recommendation methods used in business process modeling. Section 3describes the current state of the art in this research area. Selected machine learningmethods that can be used for recommending features of process models are presentedin Section 4. Section 5 presents an example which can be considered as a suitable casestudy for recommendation purposes. The paper is summarized in Section 6.? The paper is supported by the Prosecco project.

2 Types of recommendations

Basically, recommendation methods in BPs modeling can be classified as one of twotypes: subject-based and position-based classification. The first one concentrates onwhat is actually suggested, while the second one focuses on the place where the sug-gestion is to be placed. However, they are suited for different purposes and thereforeare complementary. A hierarchy of the identified types of recommendation methods ispresented in Figure 1.

2.1 Subject-based classification

In subject-based classification we focus on what is actually suggested. The suggestionitself is not directly dependent on the context it is placed in. The recommendation algo-rithms may actually inspect the context to be able to deliver more accurate results but itis not an inherent feature of recommended item.

1. Attachment recommendations – as the name suggests, these recommendationssuggest how to link a business process (or, more precisely, a selected element of it)with an external entity like a decision table or another process. Attachment recom-mendations appear naturally where user should link two already existing items.

(a) Decision tables – recommendations for a decision table describing conditionsin a gate. See an example in Figure 2.

Figure 2. Decision table suggestion

Rec

omm

enda

tion

clas

sific

atio

n

Text

ual

reco

mm

enda

tions

Posi

tion-

base

d cl

assi

ficat

ion

Stru

ctur

al

reco

mm

enda

tions

Forw

ard

reco

mm

enda

tions

Back

war

d re

com

men

datio

ns

Auto

com

plet

eN

ame

of a

n el

emen

t

Nam

e co

mpl

etio

n

Full

nam

e su

gges

tion

Serv

ice

task

Subp

roce

sses

and

ca

ll su

bpro

cess

es

Link

s

Gua

rd c

ondi

tions

Dec

isio

n ta

ble

Sing

le e

lem

ent

Stru

ctur

e of

ele

men

ts

Atta

chm

ent

reco

mm

enda

tions

Subj

ect-b

ased

cl

assi

ficat

ion

Figu

re1.

Iden

tified

type

sof

reco

mm

enda

tions

(b) Links – recommendations for a catching event that should be connected withthe selected throwing Intermediate Link Event. See an example on Figure 3.

Figure 3. Throwing Intermediate Link Event suggestion

(c) Service task – recommendation for a service task performed in the given taskitem. See an example in Figure 4.

Figure 4. Service task selection with recommendation

(d) Subprocess and call subprocess – recommendation for a subprocess or callsubprocess that should be linked with the given activity (see Figure 5).

2. Structural recommendations – a new part of the diagram is suggested. One ormore elements with, for example, missing incoming or outgoing flows are selected.The suggested structure is connected with old chosen elements.(a) Single element – a single item (activity, gate, swimlane, artifact, data object

or event) is suggested. This is a more straightforward extension of editors likeOryx/Signavio that can already insert single elements quite easily.

(b) Structure of elements – two or more items are suggested. A more sophis-ticated solution where an entire part of the process is inserted into existing,unfinished structure.

3. Textual recommendations are suggestions of names of elements or guard condi-tions. Either the full text can be suggested or suggestions may show while the textis being typed.(a) Name of an element – a name of activity, swimlane or event may be suggested.

i. Name completion happens when user is typing the name. Several possiblecompletions of partially entered name are suggested to the user.

Figure 5. Subprocess selection with recommendation

ii. Full name suggestion happens when the user wants the name to be sug-gested by the system based on the context in which the element is placed.

(b) Guard condition suggestions are different from name suggestions becausemore than one text (condition) may be suggested at once and these conditionsmust satisfy the requirements of the gateway. The latter requirement impliesthat semantic analysis of conditions is necessary to give meaningful sugges-tions. See example in Figure 6.

Figure 6. Guard condition suggestion

2.2 Position-based classification1. Forward completion – a part of the process is known and the rest of the process,

starting with one selected activity, is to be suggested. See Figure 7.

Figure 7. Forward completion

2. Backward completion – a part of the process is known and the rest of the process,ending with one selected activity, is to be suggested. See Figure 8.

Figure 8. Backward completion

3. Autocomplete – a part of the process is known and the rest of the process is tobe suggested. A number of items with no outgoing or incoming flows is selected –missing flows will lead to or from the suggested structure. See Figure 9.

Figure 9. Autocompletion

3 Recommendation Techniques for Business Process Models

Empirical studies have proven that modelers prefered to receive and use recommen-dation suggestions during design [4]. Recommendations can be based on many fac-tors, including labels of elements, current progress of modeling process, or some ad-ditional pieces of information, such as process description. There are several existingapproaches which can be assigned to the following subject-based categories:

1. Attachment recommendations: Born et al. [5] presented an approach that sup-ports modelers during modeling tasks by finding appropriate services, meaningfulto the modeler. More complex approach which helps process designers facilitatemodeling by providing them a list of related services to the current designed modelwas proposed by Nguyen et al. [6]. They capture the requested service’s compo-sition context specified by the process fragment and recommend the services thatbest match the given context. The authors also described an architecture of a recom-mender system which bases on historical usage data for web service discovery [7].

2. Structural recommendations: Mazanek et al. [8] proposed a syntax-based assis-tance in diagram editor which takes advantage of graph grammars for process mod-els. Based on this research they proposed also a sketch-based diagram editor withuser assistance based on graph transformation and graph drawing techniques [9].Hornung et al. [10] presented the idea of interpreting process descriptions as tagsand based on them provide a search interface to process models stored in a repos-itory. Koschmider and Oberweis extended this idea in [11] and presented theirrecommendation-based editor for business process modeling in [4]. The editor as-sists users by providing search functionality via a query interface for business pro-cess models or process model parts and using automatic tagging mechanism inorder to unveil the modeling intention of a user at process modeling time. An ap-proach proposed by Wieloch et al. [12] delivers a list of suggestions for possi-ble successor tasks or process fragments based on analysis of context and annota-tions of process tasks. Case based reasoning for workflow adaptation was discussedin [13]. It allows for structural adaptations of workflow instances at build time orat run time. The approach supports the designer in performing such adaptations byan automated method based on the adaptation episodes from the past. The recordedchanges can be automatically transferred to a new workflow that is in a similarsituation of change.

3. Textual recommendations: Naming strategies for individual model fragments andwhole process models was investigated in [14] They proposed an automatic namingapproach that builds on the linguistic analysis of process models from industry. Thisallows for refactoring of activity labels in business process models [15].According to Kopp et al. [16] it is not to automatically deduct concrete conditionson the sequence flows going out from the new root activity as we cannot guess theintention of the fragment designer. However, they presented how a single BPMNfragment can be completed to a BPMN process using autocompletion of modelfragments, where the types of the joins are AND, OR, and XOR.

4 Machine Learning Approach for RecommendationThe idea of recommender systems was evolving along with a rapid evolution of theInternet in mid-nineties. Methods such as collaborative filtering, content-based andknowledge-based recommendation [17] gained huge popularity in the area of web ser-vices [18] and recently most often in context-aware systems [19]. The principal rulethat most of the recommendation methods are based on, exploits an idea of similaritiesmeasures. This measures can be easily applied to items that features can be extracted(eg. book genre, price, author) and ranked according to some metrics (customer likedthe book or not). However, when applied to BPMN diagrams, common recommendersystems face a big problem of non existence of standard metrics that will allow forcomparison of models. What is more, feature extraction of the BPMN diagrams thatwill allow for precise and unambiguous description of models is very challenging and,to our knowledge, still unsolved issue.

Therefore, other machine learning methods should be investigated according to anobjective aiming at providing recommendation mechanisms for a designer. The follow-ing Section contains an analysis of possible application of machine learning methods torecommendations described in Section 2. A comprehensive summary is also providedin Table 1. The black circle denotes full support of particular machine learning methodto recommendation; half-circle denoted partial support of particular machine learningmethod to recommendation, and empty circle means no, or very limited support.

Clusteringalgorithmsa

Decisiontreesb

Bayesiannetworksc

Markovchains

Attachment recommendations m l m m

Structural recommendations m l l l

Textual recommendations m m w l

Position based classification m l l l

Table 1. Comparison of different machine learning methods for recommending features denotedin Section 2

a Useless as an individual recommendation mechanism, but can boost recommendation whencombined with other methods

b No cycles in diagramc No cycles in diagram

4.1 ClassificationClustering methods Clustering methods [20] are based on optimization task that canbe described as an minimization of a cost function that are given by the equation 1. Kdenotes number of clusters that the data set should be divided into.

N∑n=1

K∑k=1

‖Xn − µn‖2 (1)

This cost function assume existence of a function f that allows for mapping element’sfeatures into an M dimensional space of X ∈ Rm. This however requires develop-ing methods for feature extraction from BPMN diagrams, which is not trivial and still

unsolved task. Nevertheless, clustering methods can not be used directly for recommen-dation, but can be very useful with combination with other methods.

Decision trees Decision trees [21] provide a powerful classification tool that exploitsthe tree data structure to represent data. The most common approach for building a tree,assumes possibility of calculation entropy (or based on it, so-called information gain)that is given by the equation 2.

E(X) = −n∑

i=1

p(xi)log(p(xi))) (2)

To calculate the entropy, and thus to build a decision tree, only a probability p ofpresence of some features in a given element is required. For the BPMN diagram, thosefeatures could be diagram nodes (gateways, tasks, etc) represented by a distinct realnumbers. Having a great number of learning examples (diagrams previously build bythe user), it is possible to build a tree that can be used for predicting next possibleelement in the BPMN diagram. However, the nature of the tree structure requires fromBPMN diagram to not have cycles, which not always can be guaranteed.

4.2 Probabilistic Graphical Models

Probabilistic Graphical Models use a graph-based representation as the basis for com-pactly encoding a complex probability distribution over a high dimensional space [22].The most important advantage of probabilistic graphical models over methods describedin Section 4.1 is that it is possible to directly exploit the graphical representation of BPdiagrams, which can be almost immediately translated into such model.

Bayesian networks Bayesian network (BN) [23] is an acyclic graph that representsdependencies between random variables, and provide graphical representation of theprobabilistic model. The example of a Bayesian network is presented in Figure 10.

G1

B1

B2

G2

Figure 10. Bayesian network

The advantage of BN is that the output of a recommendation is a set of probabilities,allowing for ranking the suggestion from the most probable to the least probable. Forexample to calculate the probability of the value of the random variable B1 from theFigure 10, the equation 3 can be used. The G1,2 can be denoted as BPMN gateways,and B1,2 as other blocks, e.g. Tasks or Events. Thus, having any of these blocks given,we can calculate a probability of a particular block being a missing part.

P (B1) =∑G1

∑G2

∑B2

P (G1)P (B1|G1)P (B2|G1)P (G2|B1, B2) (3)

This method however, will not be efficient for large diagrams, since exact inferencein Bayesian networks is NP-hard problem. To solve this problem either the small chunksof BPMN diagram can be selected for the inference, or approximate inference applied.

Markov Chains Markov chain [24] is defined in terms of graph of state space V al(X)and a transition model τ that defines, for every state x ∈ V al(X) a next-state distri-bution over V al(X). These models are widely used for text auto-completion and textcorrection, but can be easily extended to cope with other problems such as Structuralrecommendations, or position-based classification.

We can assume different BPMN block types as states x , and connections betweenthem as a transition model τ .

Gateway Task

Event

0.25

0.75

0.4

0.3

0.3

0.5

0.5

Figure 11. Markov Chain

An example Markov chain model is presented in Figure 11. The number above thearrows denotes transition probability from one state to another. The characteristic of theMarkov chains allows for cycles.

5 Case Study

The presented recommendation approaches can be applied to process models, especiallymodeled on the basis of the existing processes in a model repository. For the purposeof evaluation, we prepared 3 different BPMN models of bug tracking systems (Djangoand JIRA) and the model of the issue tracking approach in VersionOne. A bug trackingsystem is a software application that helps in tracking and documenting the reportedsoftware bugs (or other software issues in a more general case). Such a system is oftenintegrated with other software project management applications.

Report

er

Core

Develo

per

Core Developer

Make DesignDecision Accept

Review patch

Merge andresolve as fixed

[FIXED]

Resolve aswontfix

[WONTFIX]

Postpone

Report

er

Reporter

Create Issue[UNREVIEWED]

Any C

ontr

ibuto

r

Any Contributor

Close with anyflag

[CLOSED]

Accept[ACCEPTED] Create patch Review patch

Review

Cannot bereviewed byauthor

Unreviewed

patch has issues

patch has issues

Figure 12. Django

Develo

per

Develo

per

Developer

Assessresolution

[OPEN]

Close defect[CLOSED]

Resolve defect[RESOLVED]

Prepare solution[IN PROGRESS]

QA

QA

Create defect

Testresolution

Close defect[CLOSED]

Reopen defect[REOPENED]

fixed

invalid

start progress

Figure 13. Jira

QA

, B

A, Le

ader

QA

, B

A, Le

ader QA, BA, Leader

Create defect[OPEN]

Leader

Leader

Assigndefect

Move defect tofuture iteration

[FUTURE]

Develo

per

Developer

Prepare defectsolution

[IN PROGRESS]

Commitdefect

solution[DONE]

Assesresolution

Point at buildnumber with

resolution[PENDING

TEST]

Prepare for workon defect[PENDINGANALYSIS]

Analyze defectcause

[IN ANALYSIS]

Removeblockages[BLOCKED]

QA

QA

Test resolution[IN TEST]

Accept[ACCEPTED]

Removeblockages[BLOCKED]

Reject[REJECTED]

Close defect[CLOSED]

start analysys

blocking issues

quick fix available

As designed

tests failed

blocking issues

Defect still occurs

Figure 14. VersionOne

We selected such models as a case study because of their similarity. As the processesof different bug trackers present the existing variability, such example can be easily usedto present for recommendation purposes when modeling a new bug tracking flow fora bucktracking system.

6 Conclusion and future workThis paper focuses on a problem of recommendation methods in BP modeling. Suchmethods help in speeding up modeling process and producing less error prone mod-els than modeling from scratch. The original contribution of the paper is introducinga categorization of recommendation approaches in BP modeling and short overview ofmachine learning methods corresponding to the presented recommendations.

Our future work will focus on specifying recommendation approach for companymanagement systems in order to enhance modeling process and evaluation of the se-lected recommendation methods. We plan to carry out a set of experiments aiming attesting recommendation approaches on various model sets.

References

1. OMG: Business Process Model and Notation (BPMN): Version 2.0 specification. TechnicalReport formal/2011-01-03, Object Management Group (2011)

2. Silver, B.: BPMN Method and Style. Cody-Cassidy Press (2009)3. Mendling, J., Reijers, H.A., van der Aalst, W.M.P.: Seven process modeling guidelines

(7pmg). Information & Software Technology 52 (2010) 127–1364. Koschmider, A., Hornung, T., Oberweis, A.: Recommendation-based editor for business

process modeling. Data & Knowledge Engineering 70 (2011) 483 – 5035. Born, M., Brelage, C., Markovic, I., Pfeiffer, D., Weber, I.: Auto-completion for executable

business process models. In Ardagna, D., Mecella, M., Yang, J., eds.: Business ProcessManagement Workshops. Volume 17 of Lecture Notes in Business Information Processing.Springer Berlin Heidelberg (2009) 510–515

6. Chan, N., Gaaloul, W., Tata, S.: Context-based service recommendation for assisting busi-ness process design. In Huemer, C., Setzer, T., eds.: E-Commerce and Web Technologies.Volume 85 of Lecture Notes in Business Information Processing. Springer Berlin Heidelberg(2011) 39–51

7. Chan, N., Gaaloul, W., Tata, S.: A recommender system based on historical usage data forweb service discovery. Service Oriented Computing and Applications 6 (2012) 51–63

8. Mazanek, S., Minas, M.: Business process models as a showcase for syntax-based assistancein diagram editors. In: Proceedings of the 12th International Conference on Model DrivenEngineering Languages and Systems. MODELS ’09, Berlin, Heidelberg, Springer-Verlag(2009) 322–336

9. Mazanek, S., Rutetzki, C., Minas, M.: Sketch-based diagram editors with user assistancebased on graph transformation and graph drawing techniques. In de Lara, J., Varro, D., eds.:Proceedings of the Fourth International Workshop on Graph-Based Tools (GraBaTs 2010),University of Twente, Enschede, The Netherlands, September 28, 2010. Satellite event ofICGT’10. Volume 32 of Electronic Communications of the EASST. (2010)

10. Hornung, T., Koschmider, A., Lausen, G.: Recommendation based process modeling sup-port: Method and user experience. In: Proceedings of the 27th International Conference onConceptual Modeling. ER ’08, Berlin, Heidelberg, Springer-Verlag (2008) 265–278

11. Koschmider, A., Oberweis, A.: Designing business processes with a recommendation-basededitor. In Brocke, J., Rosemann, M., eds.: Handbook on Business Process Management 1.International Handbooks on Information Systems. Springer Berlin Heidelberg (2010) 299–312

12. Wieloch, K., Filipowska, A., Kaczmarek, M.: Autocompletion for business process mod-elling. In Abramowicz, W., Maciaszek, L., Wecel, K., eds.: Business Information SystemsWorkshops. Volume 97 of Lecture Notes in Business Information Processing. SpringerBerlin Heidelberg (2011) 30–40

13. Minor, M., Bergmann, R., Görg, S., Walter, K.: Towards case-based adaptation of workflows.In Bichindaritz, I., Montani, S., eds.: ICCBR. Volume 6176 of Lecture Notes in ComputerScience., Springer (2010) 421–435

14. Leopold, H., Mendling, J., Reijers, H.A.: On the automatic labeling of process models. InMouratidis, H., Rolland, C., eds.: Advanced Information Systems Engineering. Volume 6741of Lecture Notes in Computer Science. Springer Berlin Heidelberg (2011) 512–520

15. Leopold, H., Smirnov, S., Mendling, J.: On the refactoring of activity labels in businessprocess models. Information Systems 37 (2012) 443–459

16. Kopp, O., Leymann, F., Schumm, D., Unger, T.: On bpmn process fragment auto-completion.In Eichhorn, D., Koschmider, A., Zhang, H., eds.: Services und ihre Komposition. Proceed-ings of the 3rd Central-European Workshop on Services and their Composition, ZEUS 2011,Karlsruhe, Germany, February 21/22. Volume 705 of CEUR Workshop Proceedings., CEUR(2011) 58–64

17. Jannach, D., Zanker, M., Felfernig, A., Friedrich, G.: Recommender Systems An Introduc-tion. Cambridge University Press (2011)

18. Wei, K., Huang, J., Fu, S.: A survey of e-commerce recommender systems. In: ServiceSystems and Service Management, 2007 International Conference on. (2007) 1–5

19. Bobek, S., Nalepa, G.J.: Overview of context-aware reasoning solutions for mobile devices.proposal of a rule-based approach. Computer Science and Information Systems (2014) ac-cepted.

20. Bishop, C.M.: Pattern Recognition and Machine Learning (Information Science and Statis-tics). Springer-Verlag New York, Inc., Secaucus, NJ, USA (2006)

21. Mitchell, T.M.: Machine Learning. MIT Press and The McGraw-Hill companies, Inc. (1997)22. Koller, D., Friedman, N.: Probabilistic Graphical Models: Principles and Techniques. MIT

Press (2009)23. Friedman, N., Geiger, D., Goldszmidt, M.: Bayesian network classifiers. Machine Learning

29 (1997) 131–16324. Rabiner, L., Juang, B.H.: An introduction to hidden markov models. ASSP Magazine, IEEE

3 (1986) 4–16